IRIS
by Reckoning Machines
We help investment organizations learn.
The model comes first. Everything else compounds around it.
IRIS is a model-based research operating system which preserves analytical reasoning, institutional playbooks, and organizational learning so each research cycle leaves the next better informed.
Research velocity. Institutional learning. Investment success.
Analyst pain points
Problems IRIS Solves
Context & memory
"What were we thinking going into earnings?"
"I have to paste my numbers into ChatGPT and explain the model every time."
"Where did this number come from?"
"What changed since last quarter?"
"I need the model we used for the Board meeting."
Model control
"I'm in a meeting. I don't have my model."
"I just saved over the wrong model."
"Can someone send me the latest model?"
"I have no idea who changed this formula."
"I want to schedule my model to update every morning before I get in."
Research at scale
"I changed the model. Now I have to rewrite the note."
"I do the same 5 things every time I update a model."
"Our research feels like it doesn't compound."
"I wish I could save this prompt with the model."
IRIS preserves:
Methods - how your firm reasons.
Playbooks - how your firm works.
Patterns - what your firm does repeatedly.
Observations - what actually happened.
IRIS rewards usage with retained knowledge and better process.
Figure 1 The Research Velocity Flywheel The operating model behind IRIS.
IRIS reduces the time between a question and a decision. Each research cycle preserves the context, evidence, and judgment required for the next one.
Start with the model.
Search is the front door. Open a company, model, question, or research surface without managing files.
Work in the model.
Keep the familiar grid and familiar formulas, but attach assumptions, outputs, evidence, formulas, and reasoning to the model itself.
The spreadsheet stays familiar. The context becomes durable.
METHOD
Preserve how your team reasons.
A Method captures reusable reasoning for a specific analytical judgment.
It starts in the model when an =AI() formula expresses a judgment worth reusing - though not every formula needs one.
Analysts refine that reasoning in one living document, test it against real evidence, and preserve each approved version as it evolves.
Ask the model.
Research Rail is the analyst interface to the repository. When Research Rail suggests a change, it shows you exactly what will change before anything is applied. Once you approve it, IRIS applies that exact change and records it in version history.
- Why?
- What changed?
- What stands out?
- Have we believed this before?
- Remember this.
- What if margin were 5%?
- Write the report.
Ask questions. Run research. Test scenarios. Build memory.
Take the model into the room.
Meeting Mode puts the live model in your hand. From your phone, change assumptions, re-run the model in real time, and see the conclusion and key outputs update while the conversation is still happening.
Drive your model from the meeting.
Turn work into communication.
Reports have formulas, they are built dynamically. The explanation stays attached to the assumptions, outputs, evidence, and changes that produced it.
Preserve what changed.
The timeline keeps institutional memory: what changed, when it changed, and why it mattered. Every accepted change becomes part of the model's history.
Serious research teams move faster when the model remembers.
One repository.
One timeline.
One Research Rail.
One model.
Everything stays connected.